The disclosed system and method provide a way to create, update, and execute dynamic goal plans. Updating a dynamic goal plan may be based on the initial sequence of actions of the goal plan as well as the corresponding states of the actions. By using a sequence to sequence model, a goal plan can still be processed when the length of the input (initial sequence of actions) differs from the length of the output (updated sequence of actions). A sequence to sequence model can determine the interdependencies between actions that can contribute to the optimal order in which actions can efficiently be performed. A single layer neural network or clustering can be used to approximate the state of a goal plan that may be capable infinite states. This approximation improves accuracy in capturing the state of a goal plan, thereby improving accuracy in predicting the future state of a system, which can help with planning (e.g., gathering resources in advance). Projects involving collaboration between virtual and/or human assistants can greatly benefit from the ability to update a dynamic goal plan in real time.
Legal claims defining the scope of protection. Each claim is shown in both the original legal language and a plain English translation.
2. The computer implemented method of claim 1, further comprising creating, by a goal plan module, the initial goal plan.
A system and method for automated goal planning in computational environments involves generating and optimizing goal plans to achieve desired outcomes. The method addresses the challenge of efficiently determining actionable steps to reach specific objectives, particularly in dynamic or complex scenarios where manual planning is impractical. The system includes a goal plan module that creates an initial goal plan, which defines a sequence of actions or steps designed to achieve a predefined goal. This initial plan is generated based on input parameters, such as constraints, resources, and environmental conditions, to ensure feasibility and effectiveness. The method further involves evaluating the initial plan against performance metrics, such as efficiency, resource utilization, or success probability, and iteratively refining the plan to optimize these metrics. The system may also incorporate feedback mechanisms to adjust the plan in real-time as conditions change. By automating the planning process, the method reduces manual effort, improves accuracy, and enhances adaptability in goal-driven applications, such as robotics, project management, or decision support systems. The invention is particularly useful in environments where rapid, data-driven planning is essential for achieving objectives efficiently.
3. The computer implemented method of claim 1, wherein the bidirectional RNN includes one of a long short-term memory (LSTM) and a gated recurrent unit (GRU).
This invention relates to natural language processing (NLP) and machine learning, specifically improving sequence-to-sequence models for tasks like translation or text generation. The core challenge addressed is enhancing the ability of neural networks to capture long-range dependencies and context in sequential data, which is critical for accurate predictions in NLP applications. The method uses a bidirectional recurrent neural network (RNN) to process input sequences. Bidirectional RNNs analyze data in both forward and backward directions, allowing the model to leverage context from both past and future elements of the sequence. To further improve performance, the bidirectional RNN incorporates either a long short-term memory (LSTM) or a gated recurrent unit (GRU) architecture. LSTMs and GRUs are specialized RNN variants designed to mitigate the vanishing gradient problem, enabling the network to retain and utilize long-term dependencies more effectively. These units employ gating mechanisms to control information flow, selectively preserving or discarding data over time. The bidirectional RNN with LSTM or GRU units processes the input sequence, generating intermediate representations that capture bidirectional context. These representations are then used to produce the desired output, such as a translated sentence or generated text. The inclusion of LSTM or GRU units enhances the model's ability to handle complex linguistic structures and dependencies, improving accuracy and coherence in the output. This approach is particularly useful in applications requiring high contextual awareness, such as machine translation, text summarization, or conversational AI.
4. The computer implemented method of claim 1, wherein converting the state of the initial goal plan into vector embeddings includes learning vector embeddings through a single layer neural network.
The invention relates to a computer-implemented method for converting a state of an initial goal plan into vector embeddings using a neural network. The method addresses the challenge of efficiently representing complex goal plans in a machine-readable format, enabling better processing and decision-making in automated systems. The initial goal plan, which may include multiple sub-goals or tasks, is transformed into a structured numerical representation through vector embeddings. These embeddings capture the semantic relationships and dependencies within the goal plan, allowing for improved analysis and optimization. The conversion process involves learning vector embeddings through a single-layer neural network. This neural network processes the input goal plan, extracting key features and mapping them to a lower-dimensional vector space. The single-layer architecture simplifies the embedding process while maintaining the necessary representational power. The resulting vector embeddings can be used for various applications, such as goal prioritization, resource allocation, or conflict resolution in automated planning systems. The method ensures that the embeddings accurately reflect the structure and intent of the original goal plan, facilitating effective decision-making in dynamic environments.
6. The computer implemented method of claim 1, wherein the initial goal plan includes a different number of actions from the updated goal plan.
A computer-implemented method for dynamic goal planning involves generating an initial goal plan comprising a sequence of actions to achieve a specified objective. The method monitors progress toward the goal and updates the plan in response to changes in conditions, such as new information, resource availability, or environmental factors. A key feature is that the updated goal plan may differ in the number of actions from the initial plan. For example, the updated plan might add, remove, or modify actions to optimize efficiency, feasibility, or success likelihood. The method may also include evaluating the effectiveness of the updated plan and further refining it as needed. This approach ensures adaptability in dynamic environments where rigid, pre-defined plans are insufficient. The system may use machine learning or rule-based logic to assess plan performance and generate adjustments. The method is applicable in fields such as robotics, autonomous systems, project management, and decision-making software, where flexibility in action sequences is critical for achieving objectives under varying conditions.
7. The computer implemented method of claim 1, wherein the order of the actions in the initial goal plan differs from the order of the actions in the updated goal plan.
This invention relates to a computer-implemented method for modifying action sequences in goal-oriented planning systems. The method addresses the challenge of adapting predefined action sequences to changing conditions or new information, ensuring that goals remain achievable despite variations in execution order. The system generates an initial goal plan consisting of a sequence of actions designed to achieve a specific objective. When conditions change or new constraints arise, the method updates the goal plan by reordering the actions while maintaining the overall objective. The updated plan may differ in action sequence from the initial plan, allowing for flexibility in execution. The method ensures that the reordered actions still lead to the desired outcome, even if the original sequence is no longer optimal or feasible. This approach is particularly useful in dynamic environments where rigid action sequences may fail, such as in robotics, automated workflows, or decision-making systems. The system dynamically adjusts the plan to accommodate real-time changes, improving adaptability and success rates in achieving goals.
9. The non-transitory computer-readable medium storing software of claim 8, wherein the instructions further cause the one or more computers to create, by a goal plan module, the initial goal plan.
The invention relates to a computer-implemented system for generating and managing goal plans, particularly in the context of automated planning and decision-making. The system addresses the challenge of efficiently creating and refining goal plans to achieve desired outcomes in dynamic environments. The software includes a goal plan module that generates an initial goal plan, which serves as a structured framework for defining objectives, tasks, and dependencies. This initial plan is then processed by other modules to optimize and adapt the plan based on real-time data, constraints, and feedback. The system may also include modules for monitoring progress, adjusting the plan, and ensuring alignment with evolving requirements. The goal plan module interacts with these components to ensure the initial plan is both feasible and adaptable, enabling the system to handle complex planning scenarios with improved efficiency and accuracy. The invention is particularly useful in applications requiring automated decision-making, such as project management, resource allocation, and strategic planning.
10. The non-transitory computer-readable medium storing software of claim 8, wherein the bidirectional RNN includes one of a long short-term memory (LSTM) and a gated recurrent unit (GRU).
The invention relates to machine learning systems for processing sequential data using bidirectional recurrent neural networks (RNNs). The problem addressed is improving the accuracy and efficiency of sequential data analysis by leveraging bidirectional RNN architectures, which process data in both forward and backward directions to capture contextual dependencies. The system includes a non-transitory computer-readable medium storing software configured to implement a bidirectional RNN. The bidirectional RNN processes input sequences by passing them through two RNN layers: one processing the sequence in the original order and the other processing it in reverse. This dual-pass approach enhances the model's ability to understand dependencies between elements in the sequence, improving performance in tasks like natural language processing, time-series forecasting, and speech recognition. The bidirectional RNN can be implemented using either a long short-term memory (LSTM) or a gated recurrent unit (GRU) architecture. LSTMs use memory cells and gating mechanisms to mitigate the vanishing gradient problem, making them effective for long sequences. GRUs, a simpler variant, also use gating but without separate memory cells, offering a balance between performance and computational efficiency. The choice between LSTM and GRU depends on the specific requirements of the application, such as sequence length and computational resources. The software stored on the medium enables the bidirectional RNN to be trained and deployed for various sequential data tasks, improving prediction accuracy and contextual understanding compared to unidirectional RNNs.
11. The non-transitory computer-readable medium storing software of claim 8, wherein converting the state of the initial goal plan into vector embeddings includes learning vector embeddings through a single layer neural network.
The invention relates to a system for generating and optimizing goal plans using machine learning techniques. The problem addressed is the need for efficient and adaptable goal planning in dynamic environments, where traditional planning methods may struggle with scalability and flexibility. The solution involves converting a goal plan into vector embeddings, which are numerical representations that capture the semantic relationships between different elements of the plan. These embeddings are generated using a single-layer neural network, which simplifies the learning process while maintaining the ability to encode complex relationships. The neural network processes the goal plan to produce embeddings that can be used for further analysis, such as optimization or comparison with other plans. This approach allows for more efficient planning and adaptation in real-time applications, such as robotics, autonomous systems, or decision-making algorithms. The use of a single-layer neural network ensures computational efficiency while still providing meaningful embeddings for the goal plan. The system can be applied in various domains where goal planning is critical, enabling more robust and adaptable decision-making processes.
13. The non-transitory computer-readable medium storing software of claim 8, wherein the initial goal plan includes a different number of actions from the updated goal plan.
The invention relates to a computer-implemented system for generating and updating goal plans, addressing the challenge of dynamically adjusting action sequences to achieve objectives in changing environments. The system stores software on a non-transitory computer-readable medium that generates an initial goal plan comprising a sequence of actions to accomplish a specified goal. The software monitors the execution of these actions and, based on real-time feedback or environmental changes, updates the goal plan to optimize the path toward the goal. A key feature is that the updated goal plan may differ in the number of actions from the initial plan, allowing for flexibility in adapting to new conditions or constraints. The system may also include mechanisms for evaluating the success of actions, identifying deviations from the plan, and incorporating new information to refine the action sequence. This approach ensures that the system remains responsive to dynamic scenarios, improving efficiency and reliability in goal achievement. The software may further integrate with other components, such as sensors or databases, to gather data for plan updates. The invention is particularly useful in applications requiring adaptive decision-making, such as robotics, autonomous systems, or task automation.
14. The non-transitory computer-readable medium storing software of claim 8, wherein the order of the actions in the initial goal plan differs from the order of the actions in the updated goal plan.
This invention relates to a computer-implemented system for generating and modifying goal plans, particularly in automated planning or decision-making processes. The system addresses the challenge of dynamically adjusting action sequences in goal-oriented tasks when new information or constraints arise, ensuring efficient and adaptive execution. The software stored on a non-transitory computer-readable medium includes instructions for generating an initial goal plan comprising a sequence of actions to achieve a specified objective. The system then evaluates the initial plan and, based on new data or changing conditions, updates the goal plan. A key feature is that the order of actions in the updated plan differs from the original sequence, allowing for optimized or alternative pathways to the goal. This modification may involve reordering actions, inserting new steps, or omitting unnecessary ones, depending on the updated context. The system may also incorporate feedback mechanisms to validate the updated plan, ensuring it remains feasible and aligned with the objective. This approach is useful in applications like robotics, logistics, or automated workflows where adaptability to real-time changes is critical. The invention improves upon prior methods by dynamically reconfiguring action sequences rather than rigidly following a predefined order, enhancing flexibility and efficiency in achieving goals.
16. The system of claim 15, wherein the instructions further cause the one or more computers to create, by a goal plan module, the initial goal plan.
A system for automated planning and execution involves generating and refining goal plans to achieve specified objectives. The system addresses the challenge of dynamically adapting plans in complex environments where initial conditions, constraints, or objectives may change. The system includes a goal plan module that creates an initial goal plan, which defines a sequence of actions or steps to achieve a desired outcome. This initial plan is generated based on predefined rules, historical data, or real-time inputs. The system further includes a plan refinement module that modifies the initial goal plan in response to new information or changing conditions, ensuring the plan remains feasible and aligned with the objectives. The refinement process may involve adjusting action sequences, reallocating resources, or incorporating new constraints. The system also includes an execution module that carries out the refined plan, monitoring progress and providing feedback to the refinement module for further adjustments. The system is designed to operate in environments where flexibility and adaptability are critical, such as robotics, logistics, or autonomous systems. The goal plan module ensures that the initial plan is logically structured and feasible, providing a foundation for subsequent refinements. The overall system enhances efficiency and reliability in achieving complex objectives by continuously optimizing the plan based on real-time data.
17. The system of claim 15, wherein the bidirectional RNN includes one of a long short-term memory (LSTM) and a gated recurrent unit (GRU).
The invention relates to a machine learning system for processing sequential data, particularly using bidirectional recurrent neural networks (RNNs) to improve accuracy in tasks like natural language processing or time-series analysis. Traditional RNNs struggle with long-term dependencies and vanishing gradients, limiting their effectiveness. The system addresses this by incorporating bidirectional processing, where data is analyzed in both forward and backward directions, capturing contextual information more effectively. Additionally, the system employs either long short-term memory (LSTM) or gated recurrent unit (GRU) architectures within the bidirectional RNN. LSTMs use memory cells and gating mechanisms to retain long-term dependencies, while GRUs simplify this with fewer parameters, offering computational efficiency. The bidirectional structure allows the system to leverage past and future context simultaneously, enhancing performance in tasks requiring sequential understanding. The invention improves over prior systems by combining bidirectional processing with advanced RNN variants, providing a robust solution for handling complex sequential data.
18. The system of claim 15, wherein converting the state of the initial goal plan into vector embeddings includes learning vector embeddings through a single layer neural network.
The invention relates to a system for processing goal plans, particularly in the context of artificial intelligence or automated planning systems. The problem addressed involves efficiently representing and transforming goal plans into a format suitable for machine learning or optimization algorithms. Traditional methods may struggle with high-dimensional or complex goal plans, leading to inefficiencies in planning and decision-making processes. The system includes a component that converts the state of an initial goal plan into vector embeddings, which are numerical representations that capture the essential features of the goal plan. These embeddings are generated using a single-layer neural network, which simplifies the learning process while maintaining the ability to encode relevant information. The neural network processes the goal plan to produce a compact, low-dimensional vector that preserves the relationships and dependencies within the original plan. This conversion enables downstream tasks such as optimization, comparison, or further machine learning-based analysis. The system may also include other components, such as a goal plan generator that creates the initial goal plan based on input parameters or constraints, and an optimization module that refines the goal plan using the vector embeddings. The single-layer neural network is trained to map the goal plan's state into a vector space where similar plans are close together, facilitating efficient retrieval or manipulation. This approach improves the scalability and performance of automated planning systems, particularly in dynamic or resource-constrained environments.
20. The system of claim 15, wherein the initial goal plan includes a different number of actions from the updated goal plan.
The invention relates to a system for dynamic goal planning and execution, addressing the challenge of adapting plans in real-time to changing conditions or objectives. The system generates an initial goal plan comprising a sequence of actions to achieve a specified objective. As conditions or requirements change, the system updates the goal plan to reflect new constraints or objectives, ensuring the plan remains feasible and optimal. The updated goal plan may differ in structure or content from the initial plan, including variations in the number of actions required to achieve the goal. The system monitors progress and dynamically adjusts the plan to maintain alignment with evolving objectives, improving efficiency and adaptability in goal-driven processes. This approach is particularly useful in applications where flexibility and responsiveness are critical, such as robotics, automation, or project management. The system ensures that the plan remains executable while accommodating changes, enhancing overall system performance and reliability.
Cooperative Patent Classification codes for this invention. Click any code to explore related patents in that topic.
February 19, 2020
December 20, 2022
Browse 5M+ US patents with plain-English claim translations and AI-generated analysis.